Announcements. Submitting Programs Upload source and executable(s) (Windows or Mac) to digital dropbox on Blackboard

Transcription

1 Now Playing: Vertex Processing: Viewing Coulibaly Amadou & Mariam from Dimanche a Bamako Released August 2, 2005 Rick Skarbez, Instructor COMP 575 September 27, 2007 Announcements Programming Assignment 1 is due TONIGHT at 11:59pm If you want to demo your program in person, I d like to do it Friday afternoon if possible Please contact me by (TODAY) to set up a time Submitting Programs Upload source and executable(s) (Windows or Mac) to digital dropbox on Blackboard blackboard.unc.edu Include a document that lists What optional components you did Instructions for use Any problems that you had, or components that do not work properly Please submit as a zip file with your name in the filename Last Time Today Presented the functions needed for lighting and shading in OpenGL Demoed some lighting and shading functions in OpenGL Briefly discussed Non-Photorealistic Rendering (NPR) Review the OpenGL pipeline Discuss viewing and how it applies to computer graphics 1

2 Geometry (Vertices) Rendering Pipeline OpenGL rendering works like an assembly line Each stage of the pipeline performs a distinct function on the data flowing by Each stage is applied to every vertex to determine its contribution to output pixels Vertex Processing Rasterizer Fragment Processing Pixels Vertex Processing The job of the vertex processing step is to take arbitrary input geometry, and turn it into something that the rasterizer can understand Input: arbitrary geometry, lighting and camera information Output: Shaded screen-space polygons Vertex Processing Vertex processing consists of: ModelView transform Projection transform Lighting divide Polygon clipping Viewport transform Vertex Processing Vertex Processing Vertex Processing Projection Transform Viewport Transform Vertices Modelview Transform Lighting Clipping & Primitive Assembly Vertex Processing We already talked about lighting We already talked about modeling transformations We talked a bit about projections and viewports We re going to do that in more detail today Clipping is for next time Viewing in OpenGL The OpenGL viewpoint acts as a virtual camera What parameters do you need to define a camera? Viewpoint (Center of Projection) View direction Field of view Film size Projection plane 2

3 Viewing Viewing requires 3 elements: Objects to be viewed A viewer with a projection surface A projection from the objects to the viewing surface Viewing Example: A real camera Objects: Whatever you re taking a picture of: landscape, people, etc. Viewer: The camera (with its film as the projection surface) Projection: Defined by the lens, maps 3D objects on to the 2D surface Viewing Example: OpenGL camera Objects: The input geometry Viewer: The OpenGL camera (with the view volume as its viewing surface ) Projection: The OpenGL projection matrix (GL_PROJECTION), maps 3D space (world coordinates) into 3D space (eye coordinates) Eventually into normalized device coordinates (NDC) Classical Viewing Classical views are based on the relationship between these 3 elements: objects, a viewer, and a projection In classical views, objects are assumed to be constructed from flat principal faces i.e. many buildings Used primarily by architects / engineers Planar Geometric Standard projections are assumed to be onto a single plane A projection can be perspective or orthographic In perspective projection, all rays converge at a single point (the center of projection, or COP) In orthographic projection, all rays are parallel vs. Parallel Viewing volume is a truncated pyramid aka frustum Orthographic Viewing volume is a box Orthographic 3

4 Taxonomy of Planar Geometric Planar Geometric Parallel / Orthographic Multiview Axonometric Oblique Multiview Orthographic Projection plane parallel to principle face Multiview simply means generating an orthographic view for multiple faces Commonly seen in 3D modeling programs Isometric Top View Isometric Dimetric Trimetric One-point Two-point Three-point Side View Front View Benefits Why use orthographic? Preserves distances and angles does not Can be used for measurements Why multiview? The main problem with orthographic projection is that it hides many surfaces Often add isometric view as well Axonometric Direction of projection is still perpendicular to the viewing plane But principle faces not parallel to it 3 different kinds: Isometric, dimetric, trimetric Classified by number of different foreshortening factors Axonometric Typically, one axis of space is drawn as the vertical (as seen here) Advantages and Advantages: Disadvantages Can see multiple faces of an object simultaneously Lines are scaled, but by a constant factor Could still be used for measurement Disadvantages: Angles not preserved Foreshortening does not depend on distance Not realistic 4

5 Oblique Projection Direction of projection is not perpendicular to the viewing plane Most general parallel projection Is this possible with a normal camera? Orthogonal So what is the matrix for an orthogonal projection? Assume we re looking down z How would you implement an axonometric projection? Orthogonal Projection + Rotation(s) How would you implement an oblique projection? Orthogonal Projection + Rotation(s) + Shear(s) Orthographic Examples How would you map an arbitrary bounding volume (nearxyz, farxyz) into the volume defined by (-1, -1, -1) and (1, 1, 1)? Vanishing Points In perspective projection, parallel lines (parallel in the scene) appear to converge to a single point This is called the vanishing point projections are distinguished by the number of vanishing points in the image One, two, or three One-point One principal face is parallel to the projection plane 5

6 Two-point One principal direction (i.e. axis) is parallel to the projection plane Three-point Nothing parallel to the projection plane Usually used when looking up at or down on buildings Classical Viewing Recap Classical viewing is not accurate Can be useful for various reasons Two main branches Parallel projection projection Movie Break! The Aeronaut Nicholas Lombardo Ringling School of Art & Design SIGGRAPH 2006 Available online: kerid=269&filmid=628 Doing in OpenGL We already know the commands to set up projection matrices: glortho(...) gluortho2d(...) gl(...) glufrustum(...) Now we ll talk a bit about what they really mean Positioning the Camera We usually assume that the camera is located at (0,0,0), looking down -z What do we do if we want it to appear somewhere else? Translate the world by the opposite of the new location Two ways to think about this: Whole object moves into the camera frame Camera moves (need to apply transforms in reverse order) 6

7 Projection Divide divide is the mechanism by which objects farther away are made to appear smaller How? Divide x = (x * d) / z y = (y * d) / z How would you implement this in a matrix? [ ] [ ] [ ] [0 0 1/d 0] d is the z location of the projection plane How does this work? Divide M * [x y z 1] T = [x y z (z/d)] T Need to normalize new point: normalize([x y z (z/d)] T ) = [(xd/z) (yd/z) d 1] This is the image of that point on the projection plane What happens if d = 0? Viewport Transform After perspective divide, we know where the vertices will map to in 2D But in normalized device coordinates Need to know where these will actually be displayed This is the viewport transform Viewport Transform Just need to translate into a different set of 2D coordinates From the rectangle defined by (-1, -1) and (1, 1) to the rectangle defined by (0, 0) and (width, height) How? Translate and scale What if the aspect ratio of the viewport is different from that of the camera? 7

8 Next Time Continuing with vertex processing Clipping Discussion of vertex shaders Assignment 2 will go out Reminder: Programming assignment 1 due TONIGHT by 11:59pm Upload to blackboard.unc.edu 8